Latency is only one of the problems. Most of these services seem to handle audio in a way where they are picking one person as the speaker, and muting or ducking everyone else's audio. Practices like cooperative overlap are technologically inhibited.